Daniela Amodei Warns AI Training Costs Hit $100M+ As Anthropic Faces IPO Dilemma

by Helena Blackwood
This is an opinion piece. Debate is welcome and encouraged.

AI Cash Burn Is Reaching The Stratosphere

Daniela Amodei dropped some serious truth bombs at the Bloomberg Technology Summit in San Francisco on June 4, 2026. She basically said that building smart models is ridiculously expensive. Training a frontier model like Claude 3 or its successors can easily cost over $100 million right now, and that number is scaling by 10x every single year. Private venture capital is great, but it has limits.

The public stock market is the only pool of money deep enough to keep these supercomputers humming.

Wall Street Is The Ultimate AI Fuel Station

At the heart of this shift lies the raw cost of computer chips. Tech companies are buying hundreds of thousands of Nvidia H100 and B200 chips, which retail for up to $40,000 each. Under this extreme pressure, these hardware bills represent a simple math problem that traditional venture capital simply cannot solve over the long haul.

Why Overshooting Your GPU Budget Will Ruin You

To manage these astronomical chip costs, Amodei highlighted a crucial strategy: keeping demand slightly higher than supply to protect cash flow. If you buy too many chips and nobody uses your software, you go broke fast. It is like running a trendy restaurant where people wait in line, whereas empty banquet halls with massive rent are bad for business.

Going Public Means Opening Your Secret Books

While listing on the stock market secures capital, it also triggers strict Securities and Exchange Commission rules requiring firms to show their inner financial workings every single quarter. This means Anthropic will have to reveal exactly how much they spend on cloud computing from partners like Amazon Web Services and Google Cloud—effectively giving competitors the ultimate playbook to copy.

The Secret Drama Behind Tech Boardroom Battles

This relentless drive for capital brings a heavy dose of boardroom drama. Anthropic was actually founded by Daniela and Dario Amodei, who left OpenAI in 2020 after disagreements regarding the safety and commercial direction of the company. Now, they are facing the exact same pressure to commercialize and scale up. There is a fierce debate in Silicon Valley right now about whether public shareholders will force Anthropic to ignore its safety-first mission in search of profit.

As reported by The Information, the financial pressure is intense, raising the ultimate question: can you keep AI safe when Wall Street is screaming for quarterly earnings growth?

Power Blog